A team of researchers from the University of Waterloo, Canada have been exploring the Apple Watch’s ECG sensor in some depth and found that the Apple Watch ECG sensor data could be used to develop a robust and accurate stress prediction tool.
Currently, Apple does not provide a Stress score feature on its Apple Watch platform. Other wearable players such as Samsung, Fitbit, and Garmin have offered a Stress monitoring feature for some time now.
Fitbit and Garmin Stress Score methodology
Fitbit Smartwatches not only use heart rate variability (HRV) to help determine a Stress score but also employ EDA sensors and skin temperature tracking to help Fitbit track physical indications of Stress.
Garmin watches, on the other hand, use a simplistic HRV derivation model to decipher the stress levels of users. The Garmin watch uses heart rate data to determine the interval between each heartbeat. Less variability between beats equates to higher stress levels, whereas an increase in variability represents less stress.
Some of these wearables will prompt the user when they detect less variability and offer a breathing exercise routine to offset stress signals.
Apple Watch ECG data for Stress Prediction
In the case of Apple Watches, this new study shows that it is definitely possible to deduce a stress score based on the ECG sensor data.
The Single lead ECG sensor on the Apple Watch has been proven to be very accurate. According to Apple, studies have shown good agreement in classifying the rhythm of the Apple Watch ECG compared to standard 12-lead ECGs, and in a clinical trial of 600 participants the ECG sensor had 99.6% specificity when classifying sinus rhythm and 98.3% sensitivity for atrial fibrillation.
This happens to be the first work that utilizes Apple Watch ECG for stress prediction. The results from this new study were published this week on Frontiers Digital Health.
Participants in this study were given an iPhone 7 with iOS 15.0 and an Apple Watch Series 6 containing an installed Apple Watch ECG app (WatchOS 8.3) for two weeks.
They were instructed to collect data 6 times during the day in approximately three-hour intervals. Before the ECG collection, participants were asked to complete a stress questionnaire on the iPhone using the app developed by the researchers.
ECG data was extracted using HealthKit and converted into a CSV format and subsequently loaded into Kubios to determine HRV.
Machine learning algorithms were developed and used against this Apple watch ECG data. Furthermore, they isolated the impacts of age on stress prediction models and looked at other variables such as the impact of gender on stress prediction models, the impact of the profession on a prediction model and, interestingly the impact of socioeconomic status on the stress prediction model.
The researchers found that “In general, the “stress” models had a high level of precision but lower recall. The “no stress” models performed generally well with a recall typically above 60%. Considering the ultra-short duration of the ECG measurements performed here compared to the standard, as well as the nature of real-life measurements, the results presented were quite promising.”
The heart acceleration (AC) and deceleration capacity (DC) were some of the most valuable HRV features included in the model, apart from the usual variables such as SDNN and other frequency domain features.
The researchers believe since the Apple Watch collects additional data such as sleep and physical activity, it should also be interesting to use ECG data with other stress-related variables, as they can complement the data and increase the models’ predictive power.
The researchers contend that a wearable device capable of continuous, real-time stress monitoring would enable individuals to respond early to changes in their mental health. Furthermore, large-scale data collection from such devices would inform public health initiatives and policies.
We do not know why Apple hasn’t yet offered a stress monitoring feature on its Watch platform given that the existing heart data and other activity data could be leveraged for this purpose. Its possible that they may unveil a more robust stress monitoring feature that leverages additional sensors such as EDA.
The other possibility could be that the company plans to unveil this with the next-generation AirPods that could theoretically read EEG signals and offer a more robust mental health monitoring platform. We have already seen some research this year where earbuds have been configured successfully to access EEG data from users.
Only Time will tell what Apple has up its sleeves when it comes to its products and associated features : )